A Real-Time Monitoring System for the Drivers using PCA and SVM

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056
Volume: 04 Issue: 06 | June -2017 www.irjet.net p-ISSN: 2395-0072
© 2017, IRJET | Impact Factor value: 5.181 | ISO 9001:2008 Certified Journal | Page 2970
A Real-Time Monitoring System for the Drivers Using PCA and SVM
Husam Khalaf Salih Juboori 1
1Depatment of Information Technology, Maharashtra Institute of Technology, Pune University, Pune, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract –In real-time scenarios, tracking and monitoring
a human behavior from a video is very active research. In
human computer interaction, recognition and monitoring of
end user’s expressions, emotions and behavior from the real-
time video streaming plays very important role. In such
systems it is required to track the dynamic changes in human
face movements quickly in order to deliver the required
response system. The one real time application is drowsiness
detection based on facial detection and expressions such as
drowsy driver detection in order to prevent the accidents on
road. Face and eyes behaviors based facial video analysis or
drowsiness detection for the drivers is the scope of this paper.
This paper presenting the methodology for face detection and
eyes close estimation in term of drowsiness and alertness
detection from a real-time video analysis. Cascade object
detector is used for face detection then applying Principal
components analysis (PCA)andsupportvectormachine(SVM)
in term of eyes close estimation. Whenever the drowsiness is
recognized it will buzz an alarm at the system.
Keywords: Real-time video, Face/Eyes detection,Drowsiness
detection, eyes close estimation, PCA, SVM.
1. INTRODUCTION
Drowsiness and non-alertness of the drivers are the major
causes of traffic accidents, especially for drivers of large
vehicles (such as cars, buses and heavy trucks) due to long
driving periods and lack of sleeping, so for these reasons
there is a need to develop a system to do a function in term
of decreasing the accidents. In this paper, we propose a
vision-baseddrowsinessand non-alertnessdetectionsystem
for the drivers monitoring, which is easy and flexible for
deployment in all kinds of vehicles. The system consists of
modules of face detection, eyes detection, eye
openness/close estimation, and drowsiness/alertness
measure percentage of eye close and facial appearance. The
main function of the drowsinessdetectionsystemsistotrack
and detect the feature points and the facial behaviours ofthe
people then find the drowsiness from those behaviours so
that the systems can be applied in many fields. There are
many purposes behind streets mishaps some of them are a
direct result of dozing or weariness. Consequently, car
producers have started actualizing innovation based
wellbeing frameworks. Some of these best in class
frameworks incorporate impact cautioning and brake
bolster, drowsiness discovery and astute night vision with
person on foot location. Numerousframeworksareintended
to work in current vehicles as checking frameworks for the
consideration of drivers while driving. Numerous new
advances are developing to meet the expanding wish for
more secure vehicles and roadways. A few frameworks
proposed to alert the driver when the level of driver
responsiveness appears to change showing exhaustion.
Fig -1: Modern System for Drowsiness Detection.
The innovation which is utilized as a part of these
frameworks works via naturally breaking down individual
driving qualities including the assessment of controlling
developments. In spite of this advance and improvement, it
is uncertain whether these frameworks work rapidly and
with high unwavering quality to ready drivers without
diverting them from the street conditions. Frameworks may
give a false discovery for once in a while, so with false
recognition, the drivers might be irritated by frameworks
that give repetitive notices.
In this paper an economical feasible monitoring and
detection system for the drowsy drivers is designed which
can be used to do the function, the key point isusinga simple
camera which is installed direct in front of the driver with a
MATLAB system as a processing system,thesystemwill able
to monitor the driver’s face and track the eyes to detect the
state of drowsiness. simple wired/wireless webcam isused.
The video is captured from webcam then processed by the
system and action will be taken is alarm the driver. Cascade
object detector from Viola-Jones algorithm is applied to
detect a driver’s face, eyes. Principal components analysis
(PCA) and support vector machine (SVM) in term of eyes
close estimation.

1.1 Related Work
Various vision-based structures have been produced for
drowsy driver detection in view of driver confront checking.
These systems use a camera mounted in front of the driver
that concentrationsclearly towards driver's faceandfocuses
on recognizing as indicated by the driver in high-assurance
face pictures. They can be organized on the camera used and
approach forfacial and eye incorporates figuring.IRcameras
and normal shading cameras have been used. On the other
hand, both elements focus based techniques and twofold
game plan approaches have been inspected for eye state
estimation.
In [2] presented a novel vision system for bus driver
monitoring. The system is designed for easy deployment by
using existing dome camera which is installed in front of the
driver and without using any extra hardware. The main
contribution of this system is a novel vision-basedsystemfor
bus driver fatigue detection which is applicable to low-
resolution face images captured from an oblique viewing
angle to the driver’s face. Experimental results of this system
show that proposed method is able to distinguish the
simulated drowsy and sleepy states from the normal state of
driving on the low-resolution images of faces and eyes
observed from an oblique viewing angle.
In [3] designed a driver fatigue detection system on a
smartphone and proposed a novel eye detection method
named progressive locating method (PLM). The aim of
designing PLM is getting rid of the restriction of statistic
method and trying to focus on the apparent and geometric
relationship within the facial region. The system is divided
into three parts: First, skin-colour model is employed to
detect the field of the driver’sface. Second,PLMenhancesthe
accuracy ofeyedetectionrate,Third,thedriver’swakefulness
state which is the oppositeof the fatigue stateismeasuredby
counting over the different combinations of states.
In [4] presented a yawning detection system utilizing
Embedded Smart Camera that perceives a yawning mouth
with a high distinguishing proofrate.Theysimplydofaceand
mouth revelations, with no following, and they have used
Viola–Jones count also they go without using techniquesthat
require a colossal get ready database of yawning in
perspectiveofclassifiers,inordertolessenthecomputational
time. The structure is segregated into three areas; the First
step is to distinguish the person's face. The second step is
mouth disclosure. Viola–Jones detector is used for face and
mouth recognizable proof which has starting at now been
completed in the OpenCV programminglibrary.Thelastwalk
of the technique is to choose yawning; the back projection
framework is used to do this limit. This limit is done by
finding the territory and histogram of the driver's mouth in
the main casing.
In [14] presented the proposed eye state detection system
which consists of four portions:facedetection,eyedetection,
iris detection based on CHT and eye state analysis. The
system will firstly detect the face. Secondly, after the face is
detected, then it will detect the eye on the face region. After
the eyes are successfully detected then the CHT will detect
circular shape of the eye which represented as an iris. Then
finally, the system will determine the status of the eye.
2. PROPOSED WORK AND METHODOLOGY
In this Section, first, an overview of the system framework is
presented, and then, methodology is described. The details
are focused on the face and eyes detection, eyes close
estimation and alarming system.
Fig -2: System Overview.
2.1 System Overview
The proposed framework stretches out by tending to the
previously mentioned deficiencies and along these lines
takes into consideration programmed and more solid
recognition of drowsiness time from facial video caughtbya
wireless/wired webcam which is connected with MATLAB
server.
The first step of the proposed motion-based drowsiness
detection system is face detection from facial video, which
has been accomplished by Viola and Jones object detection
framework using Haar-like features obtained from integral
images. Facial videos captured in real-life scenarios can be
subject to the problems of pose variation, varying levels of
brightness, and motion blur. When the intensity of these
problems increases, the face quality decreases.Alowquality
face produces erroneous results in detecting facial features
using either GFT or SDM. To solve this problem, pass the
detected face to Principal component analysis (PCA) and
support vector machine (SVM) those help to find the feature
extraction, image frames, Euclidean Distance and facial
expression of the input video. Thus, the low qualityfacescan
be discarded by employing thresholds to check whether the
face needs to be discarded.
The framework is shown in Figure 3. The system identifies
the face from a given video input caught by wireless/wired
camera which is connected to the MATLAB. The video will
process and detecting the eyes and mouth for that

recognized face. Detection of the drowsiness for that
recognized face will calculate by identifying the state ofeyes
(close or open). Once the eyes are detected as the close for
three seconds (as our assumption), the action will be taken
the sound will alarm the driver.
Fig -3: Proposed Framework.
2.2 Proposed Methodology
To design a system to monitor a status of the drivers,
following steps are taken:
Face and Eyes Detector: The Viola–Jones object detection
framework is the first objectdetection frameworktoprovide
competitive object detection rates in real-time proposed in
2001 by Paul Viola and Michael Jones. Although it can be
trained to detect a variety of object classes, it was motivated
primarily by the problem of face detection. Cascade Object
Detector API of viola-jones algorithm from MATLAB is used
to detect the face from the image. Once face is detected,
within the face to detect the eyes we use another Cascade
Object Detector for eyes. Then these extracted eyes are sent
to PCA and SVM.
Steps of the Principal Component Algorithm (PCA):
1) Get database set of images and then find mean of the
images
2) Find the difference between mean image and each of
database images
3) Find covariance matrix of the matrix obtained from step 2
for this covariance matrix.
4) Find Eigen values and Eigen vectors, and then we will find
the Eigen faces with larger Eigen values.
5) Find out weight vector using this Eigen faces.
6) For new/unknown image also the process will be echoed
from step 1to3 and then find out weight vectorfortestimage.
7) Now find Euclidian distance between weight vectors of
unknown image and database images.
8) If this distance is less than threshold then test image is
considered tobeindatabaseandhenceauthorized,otherwise
unauthorized.
In the flowchart of the system (Fig 4) the fatigue or
drowsiness will be detected based on theimageframesmean
of the trained video that is stored while stating the system.
The proposed system will check continuously for the match
of the expressions which is stored in the database and
according to the image frames the drowsiness will be
detected, instantly the alarm will buzz and alert will be
generated.
Fig -4: Flowchart of the system.
3. EXPERIMENTAL RESULT
The result of the proposed system based on the real-time
image frames those caught by wiredwireless webcam. As
soon as the driver is detected as a drowsy, the alarm sound
will buzz at the system. The proposed framework includes
PCA and SVM. Based on the detection of eyes and the input
datasets,stateof the drowsydriver and non-drowsydriveris
determined. The result of the proposed systemisshownhigh
accuracy in practical experiment with the actual background
and different people’s faces, table 1. Shows the experiment
results and accuracy achieved.

Figure8. shows the normal status of the driver while driving,
so that whenever ever the eyes are open,thesystemconsider
that the driver is in normal state. In the other handwhenever
the eyes are close for a while (three seconds in our case) the
system will consider that the driver is sleepy or drowsy,
Figure 9. shows the drowsy status.
Fig -5: First Background.
Fig -6: Start Webcam.
Fig -7: Train the Databases.
Fig -8: Normal Status Detected.
Fig -9: Drowsy Status Detected.
Table -1: Experiment results
Serial
number
Threshold Euclidean
distance_
min
Execution
Time
A/Q no. of
frames
1. 15 3.8948e+14 41
2. 35 3.8642e+14 43
3. 65 3.6847e+14 37
4. CONCLUSIONS
In this paper, we presented a system for a monitoring the
drivers state and drowsiness detection. The proposed
framework is using a real-time facial video. The proposed
framework begins with recognition of face from a captured
video using Viola-Jones algorithm then applied the same
algorithm for eyes, then applying PCA and SVM in term of
detection of the drowsiness. This framework is produced to

monitor the status of the drivers in term of preventing them
from the sleeping while driving. High accuracy achieved as
compare to ground truth datasets. The most restrict can be
faced this framework is the quality of the video due to
different light conditions and motions blur while driving. In
future work, the system will be extended to do the function
as independent system with the help of intelligenthardware
such as; Raspberry Pi or Arduino with more accuracy and
without the need for computer intervention during
implementation.
ACKNOWLEDGEMENT
This work is done as a first stage of the master degree
project for the first author who is pursuing a last year of
master degree in Information Technology fromMaharashtra
Institute of Technology, Pune University, Pune, India. This
work is under guidance of Assistant ProfessorLalit Kulkarni.
REFERENCES
[1] Mohammad A. Haque, Ramin Irani, Kamal Nasrollahi,
Thomas B. Moeslund, “Facial video-based detection of
physical fatigue for maximal muscleactivity”, TheInstitution
of Engineering and Technology journals (IET) Vol. 10 Iss.4,
2016.
[2] Bappaditya Mandal, Liyuan Li, Gang Sam Wang, and Jie
Lin, “Towards Detection of Bus Driver Fatigue Basedon
Robust Visual Analysis of Eye State”, IEEE Transactions on
intelligent Transportation System, 2016.
[3] Zibo Li, Guangmin Sun, Fan Zhang, Linan Jia1, KunZheng,
and Dequn Zhao, “Smartphone-based fatigue detection
system using progressive locating method”, IET Intelligent
Transport Systems, 2015.
[4] Mona Omidyeganeh, Shervin Shirmohammadi, Shabnam
Abtahi, Aasim Khurshid, Muhammad Farhan, Jacob
Scharcanski, Behnoosh Hariri, Daniel Laroche, and Luc
Martel, “Yawning Detection Using Embedded Smart
Cameras”. IEEE Transactions on Instrumentation and
Measurements, 2016.
[5] Kwok Tai Chui, Kim Fung Tsang,HaoRanChi,Bingo Wing
Kuen Ling, and Chung Kit Wu, “An Accurate ECG Based
Transportation Safety Drowsiness Detection Scheme”, IEEE
Transactions on Industrial Informatics, 2016.
[6] Aniruddha Dey, “A Contour based Procedure for Face
Detection and Tracking from Video”, 3rd IEEE International
Conference on Recent Advances in Information Technology
2016.
[7] Jun Yu1 & Zengfu Wang, “A Video-Based Facial Motion
Tracking and Expression Recognition System”, Springer
International Publishing 2016.
[8] Le Nguyen Bao, Dac-Nhuong Le, Le Van Chung and Gia
Nhu Nguyen, “Performance Evaluation of Video-Based Face
Recognition Approaches for Online Video Contextual
Advertisement User-Oriented System”, Springer India 2016.
[9] Siti Khairuni Amalina Kamarol, Mohamed Hisham
Jaward, Jussi Parkkinen, Rajendran Parthiban,
“Spatiotemporal feature extraction for facial expression
recognition”, The Institution of Engineering and Technology
journals (IET) 2015.
[10] Mohamad-Hoseyn Sigari, Muhammad-Reza
Pourshahabi, Mohsen Soryani and Mahmood Fathy, “A
Review on Driver Face Monitoring Systems for Fatigue and
Distraction Detection”, International Journal of Advanced
Science and Technology Vol.64 (2014).
[11] Ramin Irani, Kamal NasrollahiandThomasB.Moeslund,
“Contactless Measurement of Muscles Fatigue by Tracking
Facial Feature Points in A Video”, 2014 IEEE.
[12] Guha Balakrishnan, Fredo Durand, John Guttag,
“Detecting Pulse from Head Motions in Video”, Computer
Vision Foundation 2013.
[13] Nawal Alioua, Aouatif Amine, and Mohammed Rziza,
“Driver’s Fatigue Detection Based on Yawning Extraction”,
International Journal of Vehicular TechnologyVolume2014,
Article ID 678786.
[14] Norma Latif Fitriyani, Chuan-Kai Yang and Muhammad
Syafrudin, “Real-Time Eye State Detection System Using
Haar CascadeClassifierand CircularHoughTransform”,IEEE
5th Global Conference on Consumer Electronics 2016.
BIOGRAPHIES
Husam K Salih was born in
Baghdad, Iraq1989. He receivedhis
Bachelors in Engineering of
Computer Techniques from Al-
Mammon University Collage,
Baghdad, Iraq 2014. He is pursuing
the last year of Masters in
Information Technology from
Maharashtra Institute of
Technology (MIT), PuneUniversity,
Pune, India 2017. His areas of
interest include Image processing,
Artificial Intelligence

A Real-Time Monitoring System for the Drivers using PCA and SVM

Recommended

Recommended

More Related Content

What's hot

What's hot (16)

Similar to A Real-Time Monitoring System for the Drivers using PCA and SVM

Similar to A Real-Time Monitoring System for the Drivers using PCA and SVM (20)

More from IRJET Journal

More from IRJET Journal (20)

Recently uploaded

Recently uploaded (20)

A Real-Time Monitoring System for the Drivers using PCA and SVM